Having worked in the wagering industry in the past, one of the biggest challenges I noticed was how recreational punters find themselves overwhelmed by information bloat on bookmaker platforms.
A recreational punter or a casual punter is your everyday person who just wants to have some fun in their free time, whether that is at the race venue, trackside, or watching a big event from home with a bet on the side. They usually do not care much about the price they back at or the risk involved. They just want to back whatever is running next and enjoy the instant gratification of winning and are comfortable with the idea of losing. They are also, as any operator will tell you or I have personally seen, the single largest revenue-generating group in the industry.
Yet the content experience most of them get is built for someone else entirely. It is built for the form student who wants raw data (as if casual punters got time nor care for it). It is built for the very crowd the bookmakers actually dont want any business to begin with (the big red dots as we’ve marked them back in the day).
Outcome of this for your casual punter is walls of stats, tip lists that explain nothing, and previews written for page views rather than actual clarity. As Ironic as it might sound I used to generate such comments (first off the barrier, hard to overlook, look no further, …) For a recreational punter who just wants a quick steer, it is a lot of effort for very little reward.
The Core Idea
This is where Punting Buddy comes in. Punting Buddy was built to solve one thing. Make racing analysis feel like chatting with a mate. Not a dashboard, not a bloated tip sheet or robotic model dump. Just a sharp, honest punting mate.
The kind of mate who says things like “this one is the obvious pick”, “this race looks messy”, “I would leave this alone”, or “I can see why you like that one, but here is the risk”. That tone matters. Because betting is not just about data. It is also about confidence, clarity, and the enjoyment of the process. When the experience is simpler, the whole journey gets better.
Why OpenClaw Matters
OpenClaw is an open-source AI assistant that runs locally on your machine and works through the chat apps you already use, including WhatsApp, Telegram and more. It is currently used to do many things like run background tasks, browse the web, and so much more.
Skills and plugins extend what it can do, and ClawHub at clawhub.ai is where those skills are published and shared.
That is exactly what makes it the right foundation for something like Punting Buddy. It makes it possible to build an agent that feels less like a search box and more like a proper companion. Instead of dumping a wall of information on someone who just wants a fast read before the next race, it turns racing analysis in to a conversation. It is way more fun, just trust me on this.
What Punting Buddy Does
Punting Buddy is a horse racing skill built for OpenClaw. It is designed for the questions punters actually ask: what races are next, what is on today, talk me through this race, who looks the safest, who is the lively one, I fancy this horse and am I missing anything, what happened in that race.
The goal is not to pretend it has magic powers. The goal is to make race discussion clearer. It gives short conversational replies, explains why a horse appeals rather than just dropping names, and stays honest when a race looks weak or the signal is thin.
You can find the published skill here:
https://clawhub.ai/anupa-perera/punting-buddy
How I Built It
The architecture is simple on purpose, and that matters. Simple systems are easier to trust, easier to maintain, and easier to improve.
The flow is straightforward. A user asks a racing question, OpenClaw routes it into the skill, the skill works out what the user actually wants, then fetches only the minimum live data needed. The racecard acts as the backbone, and the answer comes back in a conversational punting-mate style.
A few things shaped the design decisions during the build. The Racing API free tier is rate limited to one request per second, so the skill was built conservatively: serialized requests, backoff on failures, and fewer larger fetches rather than many small ones. The skill also supports inbound racing screenshots, so a user can drop an image into the chat and it gets interpreted as supporting context rather than ignored.
Design principles
- answer the exact question first
- keep replies short by default
- fetch only the minimum data needed
- use live racecard data as the base truth
- use screenshots only as supporting context
- explain picks in plain words
- be honest when the signal is weak
Data source
For v1, the skill uses The Racing API free plan as source A. That gives access to courses and regions, free racecards for today and tomorrow, and today’s free results. That is enough to support race discovery, racecard chat, runner comparisons, quick shortlists, and results checks.
Output style
The skill is not meant to sound like an analyst writing a report. It is meant to sound like a sharp mate talking racing. That means no jargon for the sake of it, no giant content dumps, no overclaiming, no pretending weak signals are strong, and no long explanations unless the punter specifically asks for one.
Why This Matters for Punters
Recreational or casual punters are not sitting down to do homework. They want to enjoy themselves. A quick conversation before the next race, a simple steer on who to back, an honest conversation prior to the race. That is the experience they are after, and it is almost nowhere in the current industry offering.
Punting Buddy does not try to give casual punters a professional edge. It tries to match the way they already think and talk about racing. Less friction, less effort, more fun. For the punter who just wants to back something and enjoy the race, that alone is a meaningful improvement over what most platforms currently offer (especially for punters who want to have a chat and enjoy the event).
The Bigger Industry Problem
Even though this technology exists now, most of the industry has not caught up. That is the blunt truth. A lot of wagering and racing platforms are still not built for agentic integrations, and they are still thinking in old product shapes such as web apps or mobile apps.
Here are some of the biggest shortcomings:
- Low awareness of the technology : Many operators still do not really understand what agentic workflows can do.
- Little or no API support for agent use : Even where APIs exist, they are often not designed for agent-driven usage.
- No MCP support : Very few providers are thinking about MCP-style integration paths for agents and tools.
- No x402-style compliance thinking : There is still very little readiness for agent-native payment, access, and compliance flows.
- No clear distinction between agents and bots : Too many systems treat useful agents the same as spammy bots or crude scripts.
- Legacy anti-automation thinking : Some providers see all automation as a threat instead of seeing where trusted agent experiences can improve customer value and bring in more revenue.
- Poor real-time integration standards : Access to race data, betting context, and user intent is often fragmented.
- Content-first product design : A lot of the industry still optimises for content volume instead of decision quality.
- Legacy product mindset : Most bookmakers have constrained themselves to web and mobile apps as the only valid product surfaces.
The technology is moving faster than the product thinking, and that holds the whole space back.
Where This Could Go Next
The exciting part is that this is still early, but technology is pacing at an unprecedented rate. A better future is not hard to imagine.
The real opportunity for bookmakers is to start building natively for personal assistant platforms like OpenClaw. That is where the punter already lives. Everything runs through a single conversational interface they are already comfortable with. The vision is simple: ask what looks good in the next race, get a clear answer, and place the bet in the same conversation. No switching apps. No hunting for the market. For a recreational punter chasing whatever is running next, that reduction in friction is genuinely valuable.
As I see it, the era of product thinking built around web apps and mobile apps is coming to an end. Much sooner than later. If you are not already agent compliant in your product offering, be ready to be outpaced by your competition. No matter how dominant a player you are in the industry, you won’t survive if you can’t outpace the rate of delivery utilizing AI.
The security and compliance side of this is also more achievable than people assume. Tools like NVIDIA NeMo Guardrails give developers a solid foundation for building safe, policy-compliant AI agents. Responsible gambling limits, underage access controls, content boundaries. The infrastructure exists. What is still missing is the product will from operators to actually build for it.
Final Thought
Punting Buddy is a small example of a bigger idea. Recreational punters are not asking for much. They want to know what to back, get a quick reason why, and get back to enjoying the day. They do not want to become experts. They just want something that feels useful in the moment, built for the way they actually experience racing.
That is what OpenClaw makes possible here: a racing agent that feels less like software and more like a mate. And for the biggest punting demographic out there, the industry is long overdue in giving them something that actually speaks their language.